g2 conflicts

[Andy Le Blanc]

I found a statement recently that the g2 voltage should be kept slightly
positive to the plate to keep the virtual cathode in its correct position
relative to the overall potential gradients in a beam tube.

this flies against statements by O.H. Schade, allowing "an unusual flexibility of operating conditions"

Other generalized design statement that emerges when dealing with pentodes
is that the "overbiased operation of push-pull pentodes in fact, necessary"

I've seen both overbias and slightly positive g2 in fender's...

Another statement that comes out repeatedly, is that :
"screen voltages,where lower than the plate voltage, should not be obtained
by means of a series resistor but from a potentiometer across the HT line
to chassis, adequately bypassed to AF signals."

Makes me think of a tube eating marshall... assumptions will get you every time.

[Firestorm]

O.H. Schade wrote largely for people who were designing tubes, so some of his statements might be theoretical. The concern with keeping the virtual cathode in its "correct" position seems aimed at suppressing the effects of secondary emission. Since this is likely a bigger issue with pentodes than beam power tubes, it is very odd that early Fenders chose to run their screens hot.

[Gaz]

i'm just not sure if any of those statements about screen grids apply to typical guitar amps because their high voltages and how hard they drive tubes. When designing outside of a tube's rating, I think you can throw the theory out the window. I mean, every amp I've seen where the screen is positive to the plate, the screen grids begin to glow when driven - there's no way to prove that condition technically superior. I know you already know this stuff, but trust your instinct, Andy!

[Cliff Schecht]

I can't see why we would want the screen to ever be positive with respect to the plate TBH and I think you found similar results with your testing Andy. While I respect the shit out of Schade and his writing, I tend to ignore the typical biasing rules in favor for my method (which was "developed" specifically for guitar amp purposes). Putting G2 too positive I think does encourage it to start acting like a plate which it is not meant to do, it is meant to reduce the effects of secondary emissions as Firestorm says. The supressor grid can't really do it's job as effectively either if the screen is stealing some of the "bounced" electrons from the plate.

[Andy Le Blanc]

appreciate the responses...

Most of the older designs we look at were never designed to be pushed.
Finding design solutions that fit our expectations seems to be the task.

I went further down the g2 path a bit today with a voltage divider, I think Cliff
mentioned it, much better results as an amp. The bias point shifted.
I've been running 50% of rated pt ma.. The current through the tubes dropped
to 60ma from 100ma with the different g2 voltage source and had to be readjusted.

But I expected that, The real surprise was that a typical fender power side
design had slightly positive g2. Something to be aware of when you start pushing your builds.

[Gaz]

I have to mention that I worked on a JCM 900 recently that ran the screens of 6L6s more positive than it's plates by about 3-4 volts at idle. This amp had 470R screen grid resistors (stock), and dissipated about 5 watts, IIRC, on the screens running full bore.

I was really really surprised that the screen were showing no physical signs of stress, and that didn't draw any more current under these conditions. They power tubes (Winged C 6L6s) had been in the amp for 6 years to boot! Either those tubes are very rugged, or those engineers at Marshall deserve a little more credit for making that design work with a positive g2.

Andy, what was the voltage divider setup? I've though about that route using two big resistors, but was concerned about too much sag on the screens.

[Cliff Schecht]

Yeah I did the voltage divider on my SE KT88 build and really liked it a lot. That amp had a lot going on when pushed into full on distortion but the operating point didn't seem to even budge when pushed hard (which is what I wanted). The plates were at around 550V and the screens were at 350V which gave the amp lots of sparkle in the high end but still had a nice growl to it as well.

[Andy Le Blanc]

I scabbed together a couple old 40k 10w out of the bone heap.

Tap the junction between the two just to see... worked well.

If I gots the time... Today I'll do some math and see what comes of a few
chosen voltage differences. I did notice that again the operation point
changes with the different conditions across the valve.

So... there's loading, there's bias, and there's also g2 potential.

A hifi rig is not necessarily a good guitar rig, I did some schematic review
99% of guitar rigs don't take this approach, the few vintage rigs that do
treat the power side more like a SE with a larger series resistance in PS.

You can see the design shift over time, the g2 connected to the OPT ct,
the g2 connected to the OPT ct w/screen grid resistors...
the g2 connected to a separate PS tap w/screen grid resistors...

The screens change relationship when the plate goes positive with signal.
I've only found one statement to support the screen being slightly positive.

[Firestorm]

You can see the design shift over time, the g2 connected to the OPT ct,
the g2 connected to the OPT ct w/screen grid resistors...
the g2 connected to a separate PS tap w/screen grid resistors...

The screens change relationship when the plate goes positive with signal.
I've only found one statement to support the screen being slightly positive.

At least in Fenders, g2 connected to OT center tap seems to coincide with some 5D, 5E and 5F models that put a choke-based pi filter right at the front of the power supply. Although earlier, non-choke models had the screens at a lower potential connected to a separate PS node via a 10K resistor, putting the choke first sags the heck out of the power supply, so keeping a separate screen dropping resistor and cap (besides adding to the parts count) probably sounded like doo-doo.

The introduction of screen grid resistors (first 100R, then 470R) seems to coincide with moving the choke between the center tap and the screen supply AND with the introduction of a standby switch in the B+ rail. On some models this opens up one end of the choke, so when the standby switch is moved to the play position, the choke "rings" and makes a hell of a racket unless there are screen Rs to damp the surge. The 470R value seems to have stuck because it was enough to quiet that effect. And with 6V6s and 6L6s it seems to be enough to keep the screens from exceeding their dissipation rating (at least at the volumes cowboy bands played).

At no time (until much later) does it look like they manipulated g2 voltage to maximize output power or tone.

This is odd (except that they kept doing what worked) because many early tube manuals explain that keeping Vg2 below Va makes plate current largely independent of plate voltage. Although plate voltage drops as more current flows, the fixed positive screen voltage maintains electron flow to the plate as long as the electrons have enough velocity to reach it. Since the area of the screen is physically very small and the charge between the individual screen grid wires is neutralized by the virtual cathode (which extends beyond the screen per Schade) most of them do.

Everything else Fender did seems designed to maximize output (compare contemporary designs by Gibson and others) so they really missed an opportunity by failing to experiment with g2 potential. They don't make the attempt until the 135-watt models where the screens are connected in ultralinear fashion.

Andy, I don't suppose you have a high wattage pot in your parts bin; it would be fun to see the effect of a whole range of small changes in Vg2. It seems the sort of thing that could be very precisely set for different loading, bias and peak signal conditions.

[Andy Le Blanc]

I've been having the same thoughts...

I found another 40k, so the string to ground about 120k tot...

when the power side is biased at 100 ma.

its about 414v opt ct, 412v plate, 412 v after choke divider source,
356 at first tap screen supply, 355 at screen after 470r, all ref. to ground.

55v difference measured between g2 and plate.

The 412 figures at the plate and after choke do not betray the small difference
measured that got me on to it in the first place. Even after the screen resistors
the screens where .5v positive, little things....

I do have a 50r 1A ohmite pot. kicking around somewhere, might be enough.

I was also thinking of trying a simple series regulator, there's enough
current with the voltage drop to make it go, but it'll only handle 5.5w.
Might be just enough, I had one rigged for a variable plate supply to tube test.
It works, might give it a go.

That would be a good spot for a VVR, run a distortion spectrum and see
what happens at different loadings and operating points.

[Firestorm]

Hmmm. A screen-only VVR with a table of settings and their tonal effects: Andy & Dana's Sonic Cookbook.

[Andy Le Blanc]

fun to think about...

Stick to a well known loading/ opt type, say 4k fender thing, the same for pt.
You can run a variety of tube types within max V and dissipation limits.
graphically present harmonic % over a g2 v on one axis and plate dissipation on the other.

Simple graph to present data and compare actual circuits to.

before and after clip...

[Cliff Schecht]

I'm going to do some tweaking on the G2 voltage in my SE amp design thanks to Andy's great threads on screen voltage effects. I'll make sure to jot down some notes and see what kind of relationship I find.

[Andy Le Blanc]

might be on to something...

I rigged a simple tube series regulator and it worked.

It left me 60v difference between the screen and plate, plate positive.
I was able to maintain 100ma dissipation while swinging the g2 -60v to -110v
with respect to the plate. Focusing on the difference between g2 and plate.

If the dissipation figure is not maintained the dissipation is reduced as the
difference between g2 and plate is increased.

this continued until the range of the bias supply was exceeded.

G2v======dis. ma.
60=======100
65=======98
70=======96
75=======92

The numbers are demonstrative but not valid.
the wall AC is all over the place with a snow storm.

[Firestorm]

By "dissipation" do you mean plate current?

Can you rig the regulator for smaller and larger differences between plate and screen? In tetrodes and pentodes, plate current is a function of g2 voltage "within a certain range." Love to see what happens at the edges of that range. Also love to see what happens when an EL34 is run at its limits for Va and Vg2, but the difference is so big it probably does require the VVR.